I feel like your comment is in itself a great analogy for the "beware of using LLMs in human communication" argument. LLMs are in the end statistical models that regress to the mean, so they by design flatten out our communication, much like a reductionist summary does. I care about the nuance that we lose when communicating through "LLM filters", but others dont apparently.
That makes for a tough discussion unfortunately. I see a lot of value lost by having LLMs in email clients, and I dont observe the benefit; LLMs are a net time sink because I have to rewrite its output myself anyway. Proponents seem to not see any value loss, and they do observe an efficiency gain.
I am curious to see how the free market will value LLM communication. Will the lower quality, higher quantity be a net positive for job seekers sending applications or sales teams nursing leads? The way I see it either we end up in a world where eg job matching is almost completely automated, or we find an effective enough AI spam filter and we will be effectively back to square one. I hope it will be the latter, because agents negotiating job positions is bound to create more inequality, with all jobs getting filled by applicants hiring the most expensive agent.
Either way, so much compute and human capital will go wasted.
> Proponents seem to not see any value loss, and they do observe an efficiency gain.
You get to start by dumping your raw unfiltered emotions into the text box and have the AI clean it up for you.
If you're in customer support, and have to deal with dumbasses all day long who are too stupid to read the fucking instructions. I imagine being able to type that out, and then have the AI remove profanity and not insult customers to be rather cathartic. Then, substitute "read the manual" for an actually complicated to explain thing.
I am stumped. Am I misreading, or are the folks at Google deliberately confounding two interpretations of "world model"? Dont get me wrong, this is really cool, and it will undoubtedly have its use. But what I am seeing is an LLM that can generate textures to be fed into a human-coded 3d engine (the "world model" that is demonstrated), and I fail to see how that brings us closer to AGI. For AGI we need "world models" as in "belief systems". The AI model must be able to reason about (learned) dynamics, which I dont see reflected in the text or video.
>an LLM that can generate textures to be fed into a human-coded 3d engine
I'm not certain but I think the LLM is also generating the physics itself. It's generating rules based on its training data, e.g. watch a cat walk enough and you can simulate how the cat moves in the generated "world".
I see at least 2 axes here:
* Should access to a tool be restricted of it is used for malice
* Is a company complicit if its automated service is being used for malice
For 1, crowbars are generally available but knives and guns are heavily regulated in the vast majority of the world, even though both are used for murder as well as legitimate applications.
For 2, things get even more complicated. Eg if my router is hacked and participates in a botnet I am generally not liable, but if I rent out my house and the tenant turns it into a weed farm i am liable.
Liability is placed where it minimises perceived societal cost. Emphasis on perceived.
What is worse for society, limiting information access to millions of people or allowing csam, harrassment and shaming?
It is not clear that limiting Grok limits information access to millions of people, so I think your premise is flawed.
There are plenty of other resources that could serve the same people that Grok serves. Further, the fact that we aren't having discussions about ChatGPT or Claude as CSAM generators also suggests that Grok could be limited in ways that it isn't being limited currently.
It is not meant to save the doctors face. The very definition of FND is "doctors dont know what is wrong, but they acknowledge that your symptoms are real".
The point of giving it a name is in the second part. Its about explicitly acknowledging the limitations of medicine
Which when it leads to abuse it's saving face and when it's incompetence it's saving face.
For a competent doctor it's used too let a patient know they're doing their job and an acknowledgement of symptoms.
Unfortunately to a _lot_ of the field "catch-all" "diagnoses" (in intentionally separating these labels). It's the same as diagnosing someone with chronic fatigue. It's diagnosing via exclusion.
The difference between chronic fatigue and brain disorders being that you're more likely to get someone looking to make a "name for themselves" diagnosing or curing the latter vs the former...
A healthy, whole-food plant-based diet is linked to a lower risk of ischemic stroke, with studies showing reduced risk compared to meat-eaters. The conclusion of this paper[1] for example reads that "Lower risk of total stroke was observed by those who adhered to a healthful plant-based diet."
Additionally, researchers at Harvard found that a plant-based diet may lower overall stroke risk by up to 10%. [2]
Hey, do whatever helps you sleep at night. That's what I do. We're all going to the same place - a couple of years here and there won't do much when you're that old anyway.
Who would be doing the shielding? The current US government has been operating under the assumption of an incredibly expansive executive power, even over "independent" agencies.
I'm not a legal expert at all but so far the most useful mental model has been to assume absolutely no one is shielded from executive power (including organizations and people entirely outside the federal government) unless the courts have delivered a final ruling on it.
How would a final court ruling shield someone from executive power? The court relies on the executive to enforce its decisions. They can find a person in contempt, and order fines or jail time until that person complies, but I believe the orders are then enforced by U.S. Marshals, who are executive appointees.
If the Marshalls neglect their oath to the Constitution (whether ordered to by superiors or not) then the court can deputize others to carry out court orders. IIUC correctly that's usually the police.
I think there's no hard power stopping them but in general they've been reluctant to openly defy courts (they have defied lower courts some while claiming they haven't). I think at least adhering to some pretense of American democracy has so far been important to this administration and if they abandon that pretense carelessly they run a real risk of losing necessary support (popular support, Congressional support, but also support of executive branch institutions).
Also, it's not that I think they won't defy courts but they'll be very careful in doing it (well, at least as careful as this administration can be) so it's still a reasonable base assumption that court orders protect you since while it's not the ironclad protection it was it still gives you some protection for now. Though that "for now" is obviously rather ominous.
The president nominates the SEC chair, and can fire him.
This explains how, written just before Trump assumed power:
While he can't force Gensler to step down as a commissioner at the U.S. Securities and Exchange Commission, he can name a new interim SEC chair as soon as he's inaugurated on Jan. 20. He can also nominate a new commissioner to the Senate, which has to confirm the pick.
If you’re POTUS or closely related to POTUS with access to sweeping information about erratic tariff policies that actual do shift entire markets index funds index against, why couldn’t you?
In general I think you’re correct because the “inside” information is typically not as broad or powerful. But I don’t think we live in general times. If POTUS gave me personally a heads up about such adjustments on tariffs, after watching indexes tumble after announcing them, and knowing the opposite direction, I would dump almost everything I had access to in such stocks largely effected by reevaluation from tariffs.
Knowing major policy shifts before they happen from one of the powerful governments in the world is useful information, especially when the policies are being set by a handful of people who can limit access to knowledge escaping even more than regularly so markets don’t adjust from larger sets of insider information.
Now is it really considered “insider” trading in this case? Probably not, and this administration can get away with anything it seems.
That's the whole point. It CANNOT have been someone inside all 500 companies. It wouldn't make much sense because those companies did not do anything at the time. It was president Trump's tweet that moved the market. So it can only have been someone that new of Trump's announcement hours before he made that announcement.
In this administration they will be fired for cases Trump doesnt like. Who would protect their independence, Congress? The house members will not oppose any trump policy unless the US is in a depression due to his action
SEC is a federal agency so falls under the Executive which is controlled by Trump. Trump will fire any SEC employee who investigates the "wrong" people and install a loyalist in their place.
This is the truth, they have been recently divesting of anyone who wants to prosecute for white collar crime. The next four years will see at least one trillionaire because he/she will be able to game the system without any barriers because the sheriff is also the criminal.
I agree, at this point. There is of course voter agency in 2 years.
Of course lots of voters (specifically the ones that voted for this) are pretty happy with the way things are going. So any kind of blowback is uncertain at best.
If the population were genuinely interested in removing Trump, they could elect 60 democrats to the senate and a house majority, then impeach. But again, a healthy chunk are happy, and a lot if the rest can't vote Democrat for social / tribal reasons.
But make no mistake, he operates above the law because the people think it's OK. They alone have the power to remove him.
Not necessarily. First for soft reasons: media and tech companies spreading disinformation. Second for hard reasons: elections can be postponed for <reasons>.
Article II Section 2 of the U.S. Constitution, "the President shall be Commander in Chief of the Army and Navy of the United States, and of the Militia of the several States, when called into the actual Service of the United States."
The military repeatedly drills into trainees that they serve to defend and uphold the constitution, not the presidential orders. There are all kinds of lessons on when to disobey direct orders, when to speak up, how to speak up, what your protections are, etc. etc.
The military is not going to take action until it is absolutely necessary. Yes, the military is sworn to uphold the constitution, even over presidential rule but it's also well understood that a military coup should be avoided at all costs. Once initiated you cannot undo this. It has significant repercussions and has significant risk of starting a civil war. It is not a thing that starts one day and ends the next. It is a thing that at best last weeks, likely lasts months, but could last indefinitely.
It is also worth mentioning that the military is not a uniform and monolithic entity. If you're asking what arm is on the other side of that civil war, well it was originally part of the one the initiated the coup in the first place.
Trust me, no one wants a military coup. It should not be a tool used lightly. Do not ask for this until all other options have been exhausted
So far, their actions are compatible with both hypotheses.
It's when the orders come in to directly use military force to support unconstitutional acts that we find out, not simply when the military fails to intervene in legal disputes.
Military answers to whomever it thinks it's best to get paid. If relationship between government and the people fails so hard that collection of taxes to fund the military is threatened then the military sides either with the people and deposes the government or with the government and suppresses the unrest.
"the military" isn't a single entity. If things really got bad I suspect there would breaks all over the place as various groups (across all ranks) decide to either follow or refuse orders coming down. It would be chaos, the fact that things could get so bad that we're even talking about it is already a very bad sign.
I'm sure there's always some back and forth inside the structure of the military but since it's insular singular top down organization it's usually not visible to civilians and they only get exposed to the consensus that military eventually reaches. Since everybody in the military is armed there's very little benefit to actively fighting with each other using sizable force because it brings no one closer to ensuring getting paid. So any staunch opponents of consensus are just getting deposed or at most assassinated. Everything usually happens quite peacefully.
U.S. law requires the chairman of the Joint Chiefs of Staff to have served as the vice chairman of the Joint Chiefs of Staff, the chief of staff of the Army, the chief of naval operations, the chief of staff of the Air Force, the commandant of the Marine Corps, or the commander of a unified or specified combatant command.
Just after 2:00 a.m. eastern time this morning, the Senate confirmed Retired Air Force Lieutenant General John Dan Caine, who goes by the nickname “Razin,” for chairman of the Joint Chiefs of Staff by a vote of 60–25.
Although Caine has 34 years of military experience, he did not serve in any of the required positions.
The law provides that the president can waive the requirement if “the President determines such action is necessary in the national interest,” and he has apparently done so for Caine.
The politicization of the U.S. military by filling it with Trump loyalists is now, as Kendall writes, “indisputable.”
There is a chain of command in the military, you have to follow orders even if you disagree with them (unless they are unconstitutional). You may express your disagreement to your superiors, but you may not publicly distance yourself. She would be courtmartialed for this in most militaries, and probably in the US as well in a more serious situation.
I understand your frustration about the "I'm from Europe" comments, but you have to appreciate that those comments come from a similar frustration.
The universally shared image of the US is that they loudmouth how great they are compared to the rest of the world. So when news comes out on how the US fails so dramatically on very basic public/social services, it leaves people from outside the US stumped.
I'd be interested to see evidence supporting your assertion of 'universally shared image'.
The image of the United States is perceived not nearly so negative in the counties which I visited: Jordan, China, Vietnam.
I'll add a fourth country, which I haven't visited. The image of the United States in Israel, while complex, includes a great deal of admiration. As a country with security always top of mind, overall Israel broadly admires the United States' extensive defense capability and power. I doubt most Israelis would describe the US as 'loudmouth'.
Similar nuances could be laid out for the countries I listed above.
Many world citizens can parse media of a random football fan screaming 'USA!' from the state department's lengthy policy positions. The perceived image you describe resembles most closely to me a self reflection found often within the US.
Nonetheless, I do think our culture could stand to posture in a more reserved fashion broadly, while not being afraid to mention where appropriate...
We remain a positive world standout in a variety ways.
> Many world citizens can parse media of a random football fan screaming 'USA!' from the state department's lengthy policy positions.
Most recently, I believe, the main contributor to that image was the former president of the US, not some American at the FIFA World Cup. Trump providing nonstop material for satire and comedy shows didn't help either.
In general the international image of the US seems to be better under democrat presidents.
That article broadly matches with my personal observations as a European.
> I can say with high confidence that the image of the United States is perceived not nearly so negative in the following counties which I personally visited: Jordan, China, Vietnam.
Eh. I can't speak to Jordan or Vietnam, but in China opinions on the US are usually a toss up. US (cultural) exports are warring with opinionated news that doesn't portray the US any better than US news portrays China.
> As a country with security always top of mind, overall Israel broadly admires the United States' extensive defense capability and power. I doubt most Israelis would describe the US as 'loudmouth'.
You're right that overall Israelis have an extremely favorable opinion of the US (ignoring their Arab population of course), however I have to point out these opinions aren't mutually exclusive.
The only surprising thing about Israel is that Trump apparently was more liked there than Biden is. They still seem to think Trump is more arrogant though.
I definitely agree that the president has a huge affect on US image on the world stage. The critism of the analogy I laid out is fair. Trump arguably regularly carried himself in a way that aligns with the narrative we're discussing.
Props for citing evidence via Pew. Just FYI, that Israel poll arguably became obsolete in the weeks following the 10/7 terrorist attack. Not sure where things currently stand.
I do wish there was a bit more nuance. The discussion of favorability is complex. I was speaking specifically to the comment regarding the US being perceived as a 'loudmouth how great they are compared to the rest of the world'.
As far as Europe broadly is conencerned... I would probably agree that public perception does include
a component resembling the aforementioned perception. I'd speculate that this probably has a fair bit to do with many European nations exceeding the US in various metrics surrounding healthcare, happiness, safety, income distribution ect. Those critism are fair ofc.
I suppose the distilled point of my comment was that there is a lot of nuance to US image from nation to nation. Myself, I didn't realize just how respected the US is amongst so many nations. Even arguably adverserial nations often respect a great number of things about the US. I find it regrettable that the US domestic population often has very unfavorable views of their own nation, without giving adequate and nuanced consideration to our many accomplishments. Often less developed nations focus more on some of the bigger picture factors which are worthy of full consideration.
> I'll add a fourth country, which I haven't visited. The image of the United States in Israel, while complex, includes a great deal of admiration. As a country with security always top of mind, overall Israel broadly admires the United States' extensive defense capability and power. I doubt most Israelis would describe the US as 'loudmouth'.
I'm from Amsterdam and most people there adore the US as well. Our culture is heavily influenced by it as most people on the street now speak a curious mixture of Dutch and American English. The country became strongly neoliberal as a result too (which hurt me as a socialist personally) and even got caught up in the Trumpist outcry over things perceived as "woke". 24% voted for Wilders who is basically a mixture of Trump-style populism and fascism in the last election. The country is really messed up now.
But, the police is not militarised as it is in the US and I'm really glad for that. Even though gangs do have a lot of guns. They're not easy to come by in Holland but we have an unmonitored border with Belgium where they are freely available in criminal circles. We still have some 'common sense' in police approach, at least for now.
If you're asking how widely available SIMD is, it has been common in consumer hardware for 2 decades. To perform SIMD instructions manually you will need a compiled language that supports it, like Rust or C. But the compiler can actually implement it for you as an optimization.
What I'm getting from the article: it is a tool that advises affiliate marketeers on which affiliate links work, and which ones are wasted effort. I'm not 100% sure as well though.
I dont know what I expected. It's a 17 year old earning 600 bucks per month on an affiliate marketing tool.
Very cool! Although I would love to be proven wrong, I am still suspicious that this is actually a code sample from an existing game it picked up somewhere during training. Seeing how chatgpt struggles with basic logic I would be surprised if it can actually generate a new and workable game of logic.
The way I understand these models (and please correct me if I'm wrong) is that they predict every single word one at a time out of billions and billions of possible paths. So for the model to actually reproduce code you would either need to bait it really hard (reducing the possible paths by pushing it into a corner) or be impossibly "lucky". It can't just copy-paste something accidentally since it has no "awarenesses" of the source material the last neuron was created from.
It is trivial to encode text in neural networks, these models try hard to avoid that by punishing it in training but they still encode a lot of text word for word. The most famous example is "Fast inverse square root", it gives you the exact same thing with same comments etc.
When you say it's trivial to encode text in neural networks, what does that mean for LLMs? What makes it decide to encode certain texts or not? Isn't it just one big network of neurons?
The prompt I've seen for it to verbatim reproduce the fast inverse square root from Quake was:
// fast inverse square root
float Q_
When I ask ChatGPT to give me code for a fast inverse square root it doesn't reproduce it at all but gives me an implementation that looks completely different.
So, my original thought was that the prompt above with the characteristic Quake III Q_ naming is enough to push it into a corner where the path is reduced to just one possibility (with that path being the words in the code itself) and not that it merely copypasted the code from an encoded version of it. I.e. it still predicts it word-by-word but with only one possible way for each step. This is just be my naive take on it though but I really want to understand.
> my original thought was that the prompt above with the characteristic Quake III Q_ naming is enough to push it into a corner where the path is reduced to just one possibility
This is what people mean when they say it copy pastes things. It doesn't literally go to the source code, press ctrl-c and then ctrl-v that to you, nobody believes it did.
And the model does this a lot, as I said the reason it doesn't do that all the time is that they train it not to. And the quake code example got such a big deal that they started to hard code it to not return that, but that doesn't mean it never does that for other things, just that this particular example is now "fixed".
Alright, that makes sense. I still think it's an important distinction and just looking at this thread it certainly seems that people think ChatGPT is just copypasting things. Like if it ctrl-c/ctrl-v all this from a single tutorial on how to make a puzzle game.
The Quake code example still works in Copilot so I don't think they did anything about it, but only if I use the Q_ trick.
I think they did something specific to this example to make it no longer generate the code verbatim. A few months ago when I tried it "fast inverse square root" was enough to get the exact function with comments and no explanation. A week ago that gave me the same function with no comments and
"give me C code for calculating the fast inverse square root of a number, do not explain yourself"
again resulted in the original function, comments included. Today it generates a completely different result and interestingly opts to explain the origin of the technique despite the directive for it to not explain itself.
That makes for a tough discussion unfortunately. I see a lot of value lost by having LLMs in email clients, and I dont observe the benefit; LLMs are a net time sink because I have to rewrite its output myself anyway. Proponents seem to not see any value loss, and they do observe an efficiency gain.
I am curious to see how the free market will value LLM communication. Will the lower quality, higher quantity be a net positive for job seekers sending applications or sales teams nursing leads? The way I see it either we end up in a world where eg job matching is almost completely automated, or we find an effective enough AI spam filter and we will be effectively back to square one. I hope it will be the latter, because agents negotiating job positions is bound to create more inequality, with all jobs getting filled by applicants hiring the most expensive agent.
Either way, so much compute and human capital will go wasted.